[FEAT] OTel: option to use span links instead of parent context for fire-and-forget child runs

## Problem

fix(operator): replay a DAG's durable log in the order it was written

# Description

fix(dispatcher): tolerate transient worker stream backpressure

# Description

fix: batch scheduler fails to retry tasks when no worker found, engine rollout strands tasks dispatched but not yet received by worker

<!-- Please refer to the contribution guidelines before raising a PR. https://github.com/hatchet-dev/hatchet/blob/main/CONTRIBUTING.md -->

[BUG] Python SDK: contextmanager dependencies does not see raised exceptions

**Describe the issue**

[BUG] Python SDK runs dependency functions once per dependant

**Describe the issue**

[BUG] Go SDK clients panic on a malformed ID instead of returning an error

**Describe the issue**

[BUG] Filtered bulk cancel times out because metadata filter skips the GIN index

**Describe the issue**

[Go SDK] Durable tasks drop WithSkipIf and WithWaitFor when the workflow is registered

A durable task in the Go SDK ignores `WithSkipIf` and `WithWaitFor`. The task runs as if they were not set, and nothing reports an error.

feat(dashboard): order stream topics by last publish and show their status

# Description

[BUG] Python REST client cannot read a workflow version that uses `CANCEL_NEWEST` or `CANCEL_QUEUED_EXCEPT_*` concurrency

**Describe the issue** A workflow whose concurrency uses `ConcurrencyLimitStrategy.CANCEL_NEWEST` (or `CANCEL_QUEUED_EXCEPT_NEWEST` / `CANCEL_QUEUED_EXCEPT_OLDEST`) registers and runs normally, but the Python SDK's REST client cannot read its workflow version back. `hatchet.workflows.get_version(workflow_id)` raises a

[BUG] Go SDK: data race between wait completion and durable eviction

## Describe the issue

[BUG] Durable streams: a stalled reader blocks new subscribers, plus TypeScript SDK bugs and API gaps

**Describe the issue**

[BUG] TypeScript run() resolves with {} when a task passes scheduleTimeout

**Describe the issue** In the TypeScript SDK, `task.run()` and `ref.output` resolve with `{}` when a task passes its `scheduleTimeout`. The handler never ran and the run is recorded as `FAILED`, so the caller sees a success with empty output.

[BUG] Batch tasks: flush timing, stale group deadlines, member order and undocumented failure behavior

**Environment** - SDK: TypeScript 1.45.0 and 1.46.0 - Engine: self-hosted v0.115.1 and v0.116.2

Listing runs by an additional_metadata pair can time out on busy tenants

Listing runs filtered by workflow is slow on busy tenants when the workflow is rare or has no runs in the window. In our production cells, `GET /api/v1/stable/tenants/{tenant}/workflow-runs` with `workflow_ids` set to one rare workflow and `since` set to two days ago returns a 500 after about `30 s`. The same filter in

[FEAT] `mock_run`/`aio_mock_run` limitations (python sdk)

**Is your feature request related to a problem? Please describe.**

feat(durable-streams): Go, Python, Ruby SDKs

<!-- Please refer to the contribution guidelines before raising a PR. https://github.com/hatchet-dev/hatchet/blob/main/CONTRIBUTING.md -->

feat(serverless): endpoints tab and new API for endpoint events

# Description

[BUG] Ordered-release gap timeout evicts durable replays that are still releasing completions

**Describe the issue**

feat: add a Concurrency page to the dashboard

# Description

Single-task workflows release workflow-level concurrency during retry backoff

### Bug